Background
Critical illness requiring ICU admission occurs in upwards of 10,000 children each year in Canada [
1]. In addition to death, these children are at high risk for new morbidity, prolonged periods of rehabilitation and significant chronic disease [
2‐
5]. Vitamin D deficiency (VDD) (defined as a 25-hydroxyvitamin D concentration of < 50 nmol/L) is highly prevalent in pediatric intensive care units (PICU) around the globe, with rates ranging from 30 to 80% [
6‐
12]. Further, synthesis of data from multiple observational studies have demonstrated an association between lower vitamin D levels and organ dysfunction, health resource utilization and mortality in both critically ill children and adults [
6,
8‐
10,
12‐
27]. Based on the importance of adequate vitamin D status for the proper function of multiple organ systems central to critical illness pathophysiology, vitamin D supplementation has been hypothesized as a potentially simple, inexpensive, and safe intervention for improving outcomes in critically ill children [
28‐
30]. A role for vitamin D in critical illness has biological plausibility as there are multiple mechanisms through which deficiency could contribute to organ dysfunction, including: (i) exacerbation of critical illness related hypocalcemia [
31‐
33]; (ii) cardiovascular dysfunction indirectly through low body calcium stores and directly through vitamin D receptors (VDR) present on myocytes and endothelial cells; (iii) immune dysregulation through functional VDR present on all major immune cell types [
34‐
36]; (iv) through the role of vitamin D signaling in innate immunity [
37‐
39]; iv) exacerbation of critical illness polyneuropathy and muscular weakness [
40‐
43].
To date, there have been no interventional trials in the PICU setting and the adult literature is inconclusive with some but not all demonstrating that an enteral loading dose of cholecalciferol may reduce length of stay [
44], prevent mortality [
45] and improve long-term functional outcomes [
45‐
47]. A large adult phase III trial is presently underway in Europe, and will hopefully provide a clearer answer to this question [
48]. Findings from adult trials, however, cannot be extrapolated to pediatrics given differences in dosing, metabolism, co-morbidities, presenting diagnoses, and clinical outcomes [
49,
50]. Consequently, to determine whether vitamin D supplementation can improve outcomes in VDD critically ill children, pediatric trials are required.
Prior to undertaking a large randomized controlled trial (RCT) in the PICU setting, it is first essential to identify a dosing strategy that can safely restore vitamin D status in a timeframe relevant for critical illness. The current standard approach to vitamin D supplementation for the general pediatric population (200 to 1,000 IU/day of cholecalciferol) can take months to restore vitamin D status in otherwise healthy children who are deficient [
51]. Importantly, some studies in hospitalized and ICU patients have shown that 25(OH)D concentrations may decline during the initial days or week(s) of admission with standard supplementation [
52,
53] aggravated by interventions such as blood loss, blood transfusion, cardiopulmonary bypass (CPB), extracorporeal membrane oxygenation (ECMO), and plasma exchange [
7,
54,
55]. While the Institute of Medicine guidelines outline a daily age-based tolerable upper limit of 1,000 to 4,000 IU/day [
56], evaluation of studies using this regimen indicates many weeks and often a month or more may be required to restore vitamin D status [
51]. Loading dose therapy (vitamin D dose between 40,000 and 600,000 IU as a single administration, or divided over 2 days) [
51], may represent a more efficient approach to rapidly restore vitamin D status in critically ill children. However, there have been no studies evaluating the safety and efficacy of a loading dose regimen in critically ill children.
We conducted a prospective double-blind dose evaluation phase II pilot feasibility RCT. Our primary objective was to determine whether a weight-based enteral loading dose of cholecalciferol [
51] can rapidly normalize vitamin D status in critically ill children. Our secondary objectives were to: (i) evaluate whether the cholecalciferol loading dose, when compared with usual care, resulted in greater occurrence of vitamin D related adverse events, and (ii) evaluate the feasibility of a multicentre phase III trial by evaluating recruitment, protocol adherence, blinding, and seeking input from participants on a patient oriented outcome.
Methods
Study design
This trial is reported according to the Consolidated Standards of Reporting Trials (CONSORT) statement extension for pilot and feasibility trials (see checklist, Additional file
1) [
57,
58]. We conducted an international, multi-center, double-blind, phase II dose evaluation randomized controlled trial from January 2016 to November 2017. The study rationale, design, and protocol have been published [
59]. Ethical approval was obtained from the research ethics board at each participating site. Regulatory approval was obtained from Health Canada and the Austrian Agency for Health and Food Safety. Regulatory approval was not required in Chile. Written informed consent was obtained for all participants and, where applicable, written assent was also obtained. The study protocol was registered on clinicaltrials.gov by JDM on 25/05/2015 (NCT02452762). Results of this pilot trial will not be rolled into the subsequent phase III trial.
Participating centres
Patients were recruited from the PICU at four tertiary care centres: CHEO (Ottawa, Canada), London Health Sciences Centre (London, Canada), Medical University of Graz (Graz, Austria), and Hospital Guillermo Grant Benavente (Concepcion, Chile). Children were also recruited from the Neonatal Intensive Care Unit (NICU) at CHEO.
Participants and eligibility criteria
Vitamin D deficient children aged > 37 weeks corrected gestational age to < 18 years admitted to ICU with an anticipated ICU stay of > 48 h and who were expected to have access for clinical bloodwork seven days post-intervention were included. VDD for inclusion in this study was defined as a plasma 25(OH)D concentration < 50 nmol/L [
27]. This threshold was selected based on our systematic review which showed that a 25(OH)D concentration of < 50 nmol/L was associated with a > two-fold increase in mortality [
27]. A list of the study exclusion criteria are presented in Additional file
2. Eligible patients were identified and recruited in the PICU or NICU (CHEO only). In some centers, patients were also identified and pre-consented through the Cardiovascular Surgery Department and randomized if VDD was confirmed at the time of admission to the PICU.
Randomization
Participants were randomized using a web-based randomization system and assigned a randomization number. A computer-generated randomization list was prepared by the Ottawa Methods Centre at the Ottawa Hospital Research Institute. Patients were randomized 2:1 using random variable block sizes (2–4 patients/block). A 2:1 randomization (high dose: placebo) schema was employed because the control group is not directly pertinent for our primary objective of dose evaluation but required to assess our secondary objectives. Randomization was stratified by patient age (above or below 30 days of age) and by site to account for site-specific practice variation. All study personnel, members of the health care team and patients/families were blinded to study group assignment. To help maintain blinding, the active drug and placebo were identical in appearance, consistency, volume, taste and smell. The randomization number was provided to the site pharmacist and matched to a hard-copy randomization list to determine treatment allocation. The hard-copy randomization list was only accessible to the Ottawa Methods Centre and to the site pharmacist.
Intervention
The dosing regimen evaluated was identified through a systematic review and meta-regression of pediatric high dose vitamin D trials [
51]. Participants randomized to the experimental arm received a cholecalciferol load (Vitamin D3 (Cholecalciferol) Oral Solution 50,000 IU/mL, Euro-Pharm International Canada Inc.) at enrollment at a dose of 10,000 IU/kg (maximum 400,000 IU). Participants randomized to the control group received a placebo solution at enrollment, equivalent in volume to the appropriate dose of cholecalciferol. At the discretion of the health care team (who were blinded to treatment allocation), study participants could also receive routine or standard of care daily vitamin D administration (400–800 IU/day). With the exception of study drug administration (enteral cholecalciferol or placebo), there were no other changes to clinical management and no protocolization of care.
Research sample collection and analysis
Blood samples were collected either at the time of clinically indicated venipuncture or through existing arterial or central venous line access. If the participant did not have existing lines or planned clinical bloodwork, the study blood sample was not collected. Research blood samples were collected at screening (with additional blood collected at enrollment if the volume from the screening sample was insufficient for all planned analyses), days 1, 2, 3, 7 and hospital discharge for analysis of plasma 25(OH)D and ionized calcium. The research blood samples collected for an ionized calcium concentration were analyzed in real time (unless ionized calcium had been measured through clinical bloodwork in the preceding 24 h). The research blood samples collected to determine plasma 25(OH) concentrations were stored and analyzed in batches by the research laboratory by LC–MS/MS [
60]. These samples are referred to throughout the remainder of this manuscript as the
research samples.
Urine samples were collected at enrollment, on Day 3, 7, and at hospital discharge and analyzed in in real time for calcium:creatinine ratios.
Safety monitoring
Three serious adverse events (SAEs) were defined a priori that would be considered both unexpected and potentially related to study drug. The events were: (i) gastrointestinal bleeding (requiring blood transfusion) and perforation (requiring surgery) within 48 h of study drug administration; (ii) persistent hypercalcemia (> 24 h in the absence of parenteral calcium administration) with renal failure requiring dialysis, nephrocalcinosis, hemodynamically significant arrhythmia, cardiorespiratory arrest or death; and (iii) new or worsening hypercalciuria with nephrolithiasis, or renal failure leading to dialysis or death. Ionized calcium levels and urine calcium:creatinine ratios from the research samples described above were monitored in real time by the site investigator for persistent hypercalcemia [
59] and hypercalciuria (Additional file
4).
In addition, we used the local site clinical laboratory, or the Qualigen FastPak® system, to measure plasma 25(OHD) concentration in real-time from the last blood sample collected prior to hospital discharge (referred to as the
clinical sample). Three of four participating sites used a radio immunoassay to analyze the clinical samples, and the forth site used LC–MS/MS. The clinical sample result was reviewed by the study nephrologist (PG), who was not involved in clinical care of patients or any other study procedures. Safety procedures were protocolized, and patients with concerns or abnormal research samples were referred to nephrology or endocrinology as outlined in Additional file
4. The frequency of these events by study arm were reviewed by the Data Safety Monitoring Board after 15, 30 and 52 participants reached Day 7 of study enrollment.
Data collection
Data was collected from the time of study enrollment until 90 days or hospital discharge (whichever occurred first). At enrollment, families completed a questionnaire to help understand their interest in research on VDD in critical illness and inform primary outcome selection for a subsequent Phase III trial. More specifically, families were asked to indicate, other than mortality, the three most important outcomes for a research study evaluating rapid restoration of vitamin D status in critically ill children, and then to indicate all of the outcomes they would consider important.
Primary outcome
Our primary outcome was the proportion of critically ill children who achieved a plasma 25(OH)D concentration > 75 nmol/L (normalization of vitamin D status) prior to hospital discharge based on the
clinical samples with imputed concentrations for missing data (see Statistical Analysis for description of imputation methods). We initially intended to perform the primary analysis using the
research plasma sample collected on Day 7. However, an instrumentation issue recognized during analysis of the batched research samples resulted in inaccurate plasma 25(OH)D measurements. As there was insufficient research sample volume to repeat the laboratory analysis for all research samples, the study steering committee made the decision to evaluate the primary outcome based on the
clinical sample collected and analyzed in real-time prior to hospital discharge. We have included the research sample results in Additional file
3; the results reported below in the manuscript are based on the clinical sample.
Secondary outcomes
Our secondary outcomes included the frequency of vitamin D related adverse events and an evaluation of the feasibility of a large phase III trial. Vitamin D related adverse events were defined as: persistent hypercalcemia > 24 h without calcium administration [
59]; hypercalciuria as determined by an elevated calcium:creatinine ratio in two sequential post-intervention urine samples; and ultrasound-confirmed nephrocalcinosis. Feasibility outcomes for a multi-centre phase III RCT included protocol adherence, rate of study drop-out, ability to maintain blinding, assessment of the study eligibility criteria, and patient accrual. Criteria for feasibility and proceeding with a Phase III trial were established, and are summarized in Table
4. We also reported baseline values for two potential outcomes for a Phase III trial in order to inform sample size for subsequent phases of this research program: (i) multi-organ dysfunction (measured by the Pediatric Logistic Organ Dysfunction (PELOD-2) score [
61]) at enrollment and on Day 3, 7, 30, and PICU discharge; and (ii) PICU length of stay.
Sample size
The weight-based loading protocol was designed to achieve a 25(OH)D concentration of > 75 nmol/L in 75% of study participants in the intervention arm, with a minimal acceptable proportion of 50% achieving this target. Assuming the true proportion achieving target was 75%, a random sampling of 36 patients would have ~ 90% power to return an estimate in excess of 66% of participants achieving target 25(OH)D. Given an estimate in excess of 66% and a sample size of 36, the lower 95% confidence interval would exclude 50%. To account for an anticipated 5% drop out rate or missing blood samples, we aimed to recruit a total sample size of 60 patients: 40 patients into the high dose arm, and 20 in the placebo arm. Although a control group was not relevant for the primary outcome (evaluating response to vitamin D loading dose), it was determined to be essential for evaluating abnormalities in blood or urine calcium levels. Of these two, urine calcium levels are infrequently measured in the ICU, and having a control group would be essential to interpreting the levels. For example, our pilot randomized controlled trial of pre-operative vitamin D supplementation in stable congential heart disease demonstrated that 17% of patients have elevated urine calcium peri-operatively [
62]. Second, without the placebo arm we would not be able to properly evaluate recruitment or our ability to achieve blinding.
Statistical analysis
All statistical analyses were performed using R version 3.6.3 [
63]. The treatment and placebo arms were described separately using means with standard deviations for normally distributed continuous variables or medians with interquartile range values for non-normally distributed variables. Categorical variables were described using frequencies with percentages.
The proportion of participants in the treatment and placebo arm achieving a 25(OH)D concentration > 75 nmol/L by hospital discharge was calculated, with the Wilson’s score used to generate 95% confidence intervals. Additionally, the difference in these proportions between groups was calculated, again with a 95% Wilson confidence interval, and Fisher’s exact test for comparing the proportions. As stated above, due to an instrumentation issue during analysis of the research samples, we instead used the clinical sample analyzed in real-time prior to hospital discharge to evaluate the primary outcome. As a clinical sample for the primary analysis was not available for all patients, an imputation procedure was performed when an accurate research measurement was available. First, using observed pairs of clinical and research measurements of 25(OH)D, a linear regression log-transformed model of clinical sample measurements on log-transformed research sample measurements was fitted. This model was then used to impute the clinical sample measurement when an accurate research sample measurement was available. The results were first reported using imputation of missing 25(OH)D values and second using solely the clinical sample 25(OH)D concentrations. An analysis of covariance (ANCOVA) was performed to compare 25(OH)D concentration from the clinical sample between treatment groups, adjusting for the baseline 25(OH)D concentration at the time of screening (prior to study drug administration).
Secondary analyses included calculating proportions by treatment group of the following potential adverse effects of treatment: presence of nephrocalcinosis, persistent hypercalcemia, and persistent hypercalciuria. Wilson’s score was used to generate 95% confidence intervals for each above-mentioned proportion. As described in the trial protocol paper [
59], we also collected and reported common PICU clinical outcomes by treatment group. Statistical tests were not performed to compare the groups.
Sub-analyses were performed whereby the proportion of participants in the treatment arm and in the placebo arm achieving 25(OH)D concentrations > 75 nmol/L was calculated with Wilson’s score’s test to generate 95% confidence intervals within six different subgroupings: (1) patients under and ≥ 40 kg or over; (2) patients with a 25(OH)D concentration at the time of screening above or ≤ 32 nmol/L; (3) newborns (< 30 days old) or ≥ 30 days of age; (4) by country of enrollment, Canada versus outside of Canada; (5) by race, and (6) by admitting diagnosis. To test for subgroup differences a logistic regression model was fit with a main effect of treatment arm, the subgroup variable (e.g. weight category), and an interaction between these two variables. The p-value for the estimated interaction was reported.
Significant modifications to the study drug protocol
The trial intervention initially involved two doses of study drug (placebo or cholecalciferol): (i) at time of enrollment, and (ii) a day 3 dose dependent on the 25(OH)D concentration achieved. However, after randomization of the first 12 patients, the intervention was adjusted in July 2016 to a single-dose protocol administered at the time of study enrollment. There were multiple reasons for this modification. First, new literature in the adult ICU population demonstrated that administration of a single loading dose of 500,000 IU, comparable to our intervention, raised 25(OH)D concentrations by 80 (± 35) nmol/L [
44]. These findings suggested that incorporating a second load into the protocol might be of little value. Secondly, three of the initial twelve patients met the criteria for safety follow-up based on their discharge 25(OH)D concentration. At this time, the threshold for safety follow-up in our study protocol was a 25(OH)D concentration > 150 nmol/L, however, this was later increased to 200 nmol/L based on guidance from the Canadian Paediatric Society [
64]. Of these three patients, two had already achieved target 25(OH)D concentrations by Day 3 following a loading single dose. Thirdly, expanding recruitment to additional centres and eventual translation of the dosing regimen would be more feasible with a protocol involving a single loading dose only. Fourth, removing the second loading dose substantially decreased study costs. Finally, since the second loading dose first required determining the patient’s 25(OH)D concentration at Day 3, the chances of accidental unblinding were increased. Of the twelve patients who were randomized while the two-dose protocol was active, seven received > 1 dose of cholecalciferol and their 25(OH)D data was excluded from the analysis. A summary of results for these seven patients is presented in Additional file
5.
Discussion
This pilot phase II RCT confirms that a single 10,000 IU/kg dose can rapidly and safely normalize plasma 25(OH)D concentrations in critically ill children identified as vitamin D deficient (< 50 nmol/L) during admission to the PICU, and that proceeding with a Phase III trial is feasible.
We recruited vitamin D deficient (median ~ 35 nmol/L) children from 4 PICUs and 1 NICU at 4 academic centers in Canada, Austria and Chile. The evaluation of baseline characteristics demonstrates the study cohort was severely ill, with 80% requiring mechanical ventilation and 50% receiving vasoactive agents. The loading dose regimen rapidly increased blood levels within 1 day, peaked on days 2 and 3, with 80% of participants exceeding the target blood threshold of 75 nmol/L. These findings are similar to studies with comparable doses administered to critically ill adults with admission 25(OH)D levels < 50 nmol/L. For example, the VITdAL study reported an increase of 50 nmol/L following a 540,000 IU cholecalciferol dose, with 52% exceeding 75 nmol/L [
45]. Similarly, in the VIOLET trial, an adult acute lower respiratory tract infection population at high risk for acute respiratory distress syndrome and ICU admission increased blood 25(OH)D concentrations from an average of 28 to 117 nmol/L (75% exceeding 75 nmol/L) with a 540,000 IU enteral dose [
65]. No pediatric trials have evaluated a similar age or weight base dosing regimen on critically ill children. The most related study evaluated the 10,000 IU/kg dose recommended from our systematic review, in a cohort of VDD children undergoing cardiac surgery for congenital heart disease (Tetralogy of Fallot) [
66]. Cholecalciferol dosing approximately two weeks prior to surgery produced significantly higher concentrations in the treatment arm relative to control (83.5 vs. 27.4 nmol/L) immediately before the surgery. Given that blood 25(OH)D concentrations peaks ~ 72 h following loading dose administration and then begins to decline [
51], and the known half-life of vitamin D, blood 25(OH)D concentrations may have been higher if they had been measured closer to loading dose administration.
In addition, this study also assessed the feasibility of the weight-based loading dose regimen through an evaluation of vitamin D toxicity data. Loading dose cholecalciferol, when using excessive or repeated doses or in particular in infants and young children, has been linked to hypercalcemia, nephrocalcinosis and renal failure [
67]. Critically ill patients may be at greater risk for adverse events due severe organ dysfunction, higher prevalence of genetic abnormalities, and potential interactions with common ICU medications [
68,
69]. However, we did not document any cases of persistent hypercalcemia, consistent with the low hypercalcemia rate (2.6%) observed in a meta-analysis of pediatric interventional trials [
7]. These findings align with the hypercalcemia rates reported in the adult VITdAL-ICU (≤ 1%) and VIOLET (≤ 3%) trials, with no differences between control and intervention arms [
45,
65]. Similarly, in a pediatric cardiac surgery trialblood calcium levels in the arm receiving a 10,000 IU/kg pre-operative dose were not elevated compared to the control group [
66]. Evaluation of urine calcium and nephrocalcinosis data in our trial failed to identify reason for concern with hypercalciuria rates similar between treatment (15%) and usual care (21%). Nonetheless, the baseline rate was considerably higher than the pooled 2.5% rate reported in the systematic review of pediatric vitamin D clinical trials (no critically ill children) [
51]. This difference was not unexpected in the pediatric critical care setting due to the significant presence of systematic inflammation, renal dysfunction, and medication use (e.g. diuretics). Again, findings were very similar to those presented in the adult VITdAL-ICU trial, where both arms had 25% hypercalciuria rates both before and following a 540,000 IU cholecalciferol load [
45]. Sahu and colleagues also reported similar or potentially lower average post cardiac surgery calcium:creatine ratios in the cohort receiving the cholecalciferol load (2.0 vs 1.1, p = 0.16) [
66]. All of this suggests that hypercalciuria is not an appropriate biomarker of excess vitamin D levels in the ICU setting. Finally, further suggesting the safety of this regimen, was the absence of any definitive cases of nephrocalcinosis in the treatment arm or serious adverse events potentially related to vitamin D.
An important study observation relates to significant variability in post loading dose 25(OH)D concentrations. Eight of 10 of participants in the treatment arm exceed the 75 nmol/L threshold following the single loading dose. However, while the group average 25(OH)D concentration achieved was successful, individual patient responses were variable with some patients exhibiting minimal change and not reaching concentrations > 75 nmol/L (n = 7). In addition, a few patients achieved concentrations > 200 nmol/L (n = 4) or > 250 nmol/L (n = 2). Significant variability in post-study drug concentrations has been previously recognized, with regression analysis of data from pediatric interventional trial literature calculating the SD to average 42% of the mean 25(OH)D [
51]. The calculated SD of ~ 63 nmol/L on a mean plasma 25(OH)D concentration of ~ 126 nmol/L and the observation ~ 20% of study participants had concentrations either below 50 nmol/L or above 200 nmol/L is consistent with these previous findings. These findings are also in line with the post-loading dose SD of 52 nmol/L (day 7) and 58 nmol/L (day 3) in the VITdAL and VIOLET trials, respectively [
45,
65]. Both VITdAL (24%) and VIOLET (12%) also reported a significant number of non-responders, defined as post-drug 25(OH)D concentrations below 50 nmol/L. Comparing results with respect to elevated 25(OH)D is more challenging due to application of different thresholds across studies. In the VITdAL analyses, Amrein et al. applied a 150 nmol/L threshold, reporting 13% of study participants above this level with the two highest 25(OH)D concentrations in their treatment arm as ~ 265 nmol/L [
45], below the 375 nmol/L threshold where risk of acute toxicity may begin to rise [
70,
71]. With respect to the VIOLET trial, the upper target 25(OH)D concentration was set as 300 nmol/L, with only one patient exceeding that value [
65]. The two highest 25(OH)D measurements in our pilot study were 343 and 275 nmol/L, and with no symptoms of vitamin D toxicity. Our protocol indicated a change in dose would be considered if > 10% of the participants receiving the loading dose achieved a 25(OH)D concentration above 250 nmol/L [
59], and this was not exceeded [
70,
71]. As the first evaluation of this dosing regimen in a high-risk population, we made the decision to use a conservative threshold of 250 nmol/L. Although our aim was to avoid 25(OH)D concentrations > 250 nmol/L, we still recommend proceeding with this dosing regimen for a subsequent large-scale trial. Vitamin D toxicity is time-dependent, and transient levels > 200 nmol/L do not appear relevant [
71]. Importantly, there were no cases of persistent hypercalcemia, clinically significant hypercalciuria, nephrocalcinosis or any other adverse events related to vitamin D supplementation observed in this pilot study. Further, reducing the dose would increase the number of patients who do not achieve post-supplementation 25(OH)D concentrations > 75 nmol/L and dilute the impact of the loading dose regimen on clinical outcome in a Phase III trial. This decision was also informed by observations by our group, and others that 25(OH)D concentrations peak 72 h following loading dose administration and then rapidly begin to fall [
51], with an average decline of 10 nmol/L week observed in our systematic review of pediatric loading dose trials [
51]. For example, Thacher et al
. found that 25(OH)D3 concentrations fell by 53% and 59% in rachitic and healthy children 14 days following administration of a loading dose of 50,000 IU [
72].
The results of this phase II trial support the feasibility of large-scale, multicentre phase III trial powered for clinical outcome. We met our a priori established feasibility criteria for protocol adherence, blinding, study withdrawal and patient accrual. The patient accrual rate was 3.4 patients per month (~ 1.9 patients/month/site). However, with the exception of the lead site, the other sites were only recruiting for ≤ 5 months, while peak recruitment in critical care RCTs is generally not achieved until at least 7 months after a site initiates recruitment [
73]. Therefore, we believe that a reasonable expected accrual rate for a multi-year, multi-site RCT would be 2 patients/month/site, which is within the range observed in other completed large, multi-centre PICU RCTs performed in the last decade by large research consortiums [
74‐
81]. Of note, this expected accrual rate is dependent on a VDD prevalence consistent with that observed during this pilot study. Our ability to collect blood samples at Day 7 was slightly lower than anticipated (84% vs > 95%). In this trial, the Day 7 blood sample was essential to establish the efficacy of the dosing regimen, and to evaluate for toxicity; however, sample collection would be less important for a more pragmatic Phase III trial focused on clinical outcome. Given that 25(OH)D concentrations peak ~ 72 h following loading dose administration, adjusting the protocol to allow sample collection anytime between Day 2 to Day 7 would increase the frequency of sample collection, as clinical bloodwork frequency tends to be higher earlier during a PICU admission. This would allow a future Phase III trial to also perform a metabolomic sub-study to supplement the existing adult literature [
82‐
84].
Recognizing the importance of patient-centred outcomes in clinical trials, we engaged with participating families to seek input on the primary outcome for the Phase III trial. Feedback from families participating in this trial and a concurrent survey by Merrit et al. [
85], indicates that health-related quality of life is the most important outcome for families in a Phase III clinical trial in critically ill children [
85]. Historically, PICU RCTs have used hospital-based outcomes (e.g. length of stay in PICU, hospital mortality) that allow primary outcome data collection for 100% of enrolled patients. In contrast, previous observational studies evaluating health-related quality of life in PICU patients at ≥ 1 month have reported completion rates of 52 to 79% [
86‐
88]. Using a primary outcome that will be collected following hospital discharge for the Phase III RCT means that some participants will be lost to follow-up, and this should be accounted for during sample size calculation. Further, strategies to maximize follow-up will be essential, such as collecting multiple points of contact information, including locators; pre-notifying participants of an upcoming follow-up visit; making additional attempts to contact if the first attempt is not successful; monitoring loss to follow up to identify concerns early; and the addition of incentives for completing questionnaires [
89‐
91].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.